Towards Smart Financial Control: Leveraging Intelligent Data Analysis Techniques to Enhance the Reliability and Efficiency of Digital Banking Transfers

Abstract

This study aims to enhance the reliability of digital financial control by developing an intelligent framework (AEGIS-FD), an Autoencoder-Enhanced Gain-based Intelligent System for Fraud Detection capable of overcoming the challenges of massive data volumes and severe class imbalance in fraud detection. The study utilized a substantial sample from the Kaggle Credit Card dataset, comprising 284,807 financial transactions to ensure comprehensive testing. The research tool consists of a novel hybrid model that integrates Deep Autoencoders, as an unsupervised learning approach, with the Gaining-Sharing Knowledge(GSK)algorithm as an optimization tool for network architecture and parameters. The key findings of this study demonstrate the effectiveness of merging nature-inspired algorithms with deep learning to improve detection accuracy. The proposed model achieved a superior F1-score of 0.920 and a Recall of 0.898, with an exceptional operational efficiency of 1,100 transactions per second, confirming its viability as an advanced financial control tool for real-time banking Transfers.